First, a correction to the commonly cited “95%” claim: the original draft attributed this figure to BCG, but this review found no BCG first-party data supporting the claim that “95% of companies spend their entire budget on technology.” The 95% figure is often explained by citing MIT Project NANDA’s preliminary 2025 report, but the original download URL now redirects, and MIT’s current NANDA publications page does not provide a verifiable version. Until the denominator and definitions can be confirmed from a first-party document, this article does not treat 95% as a publishable fact or a general failure rate.
BCG’s 10/20/70 principle, presented in 2026, is a guideline for allocating effort: 10% on algorithms, 20% on technology and data, and 70% on people and processes, so that change actually takes hold. The 70% is not “a precise breakdown of AI value,” nor does it support the claim that “95% of budgets go to technology.” The practical management takeaway is to confirm processes, roles, training, and adoption mechanisms before purchasing tools.
A 2026 first-party McKinsey article, citing its global survey, notes that only 21% of companies have fundamentally redesigned their operating model around AI. High performers that attribute at least 5% of EBIT to AI are about three times as likely as others to have undertaken broad operating model redesign, and about twice as likely to redesign processes before selecting tools. These are correlations observed among surveyed companies and do not mean that redesigning processes or adopting a particular product guarantees returns.
Unima recommends replacing tool checklists with four verification gates: 1. Outcome baseline: first define the current state and owner for revenue, gross margin, processing time, error rate, or service volume. 2. End-to-end process: map bottlenecks, human judgment, data sources, and cross-department handoffs, rather than automating a single step. 3. Permissions and exceptions: define the situations in which AI may act, requires human approval, or must stop. 4. Scale-up conditions: decide whether to expand based on adoption rate, unit cost, quality, and financial impact, and keep a rollback plan. This is an implementation recommendation, not a specific client result or a return commitment.
Sources and limitations: BCG, “Scaling AI Requires New Processes, Not Just New Tools,” 2026-01-20: https://www.bcg.com/publications/2026/scaling-ai-requires-new-processes-not-just-new-tools; McKinsey, “The operating model advantage: Why AI winners are rewiring their organizations,” 2026-07-07: https://www.mckinsey.com/industries/industrials/our-insights/the-operating-model-advantage-why-ai-winners-are-rewiring-their-organizations; MIT Media Lab NANDA current publications page: https://www.media.mit.edu/groups/nanda/publications/. This article was verified on 2026-08-25; it does not use the 95% figure, which cannot be verified against current first-party documents, as a headline, outcome assessment, or sales claim.